PDA

View Full Version : Machine Accuracy Issues PRO4824 Mach3



pete1961
03-06-2014, 09:10 PM
I thought I would turn to the forum for some further expert advise!

I have a PRO4824/Nema23/mach3 Setup.

I have an issue with the router not being very accurate.
I believe it is a mach3 issue but I'll take any advise as to what to check.

When I cut circular pockets say 3-1/2" or 2-1/4" in MDF or Acrylic the end result are holes with an OD that is undersized by .020 - .023 inches.
Grooves or Slots also produce a width along the slot that is undersized by an almost similar amount.
I consider this an un-acceptable amount.

The perplexing issue is I have calibrated both X/Y axis to a distance of 4" & 5" with precision gage blocks, dial test indicator, dial indicator, calipers. Pointing out I'm doing this calibration with precision equipment. After the calibration is complete I return to the 0,0 point (Work Zero) of where the calibration started from and get a reading of being dead on at 0,0. One I jog move the axis is immediately inaccurate by .002-.005 thousands and grows somewhat larger as I move closer to the original calibration distance of the 4" or 5". The error occurs in single step and continuous mode. I figure things aren't quite right, proceed to re-calibrate mach3 for the correct steps repeating the process over again. This subsequent second time I have mach3 do a calibration move 5" in X or Y it moves perfectly that distance in calibration mode with no changes needing to be made to the motor steps.

Cory at CNC Router parts mentioned earlier today to run in CV mode rather than exact mode, upon checking I'm already running in CV mode so thought I would switch to exact mode to see if it made a difference. It made no difference. Further I'm believing the CV Mode setting relates to circular or arc movements only and as noted above it is happening with both circular cuts and straight line movement.

I've ruled this out as being a mechanical issue unless someone here thinks otherwise. I've gone over everything mechanical and find nothing out of order.

Hoping someone here can offer some guidance on what the problem could be.

jalessi
03-06-2014, 09:47 PM
Pete1961,

Which stepper driver and breakout board are you using?

Jeff...

ger21
03-06-2014, 09:49 PM
You really should be calibrating to a larger distance than 5". For all you know your machine is perfect at 5", and could be off an inch at 24".
Personally, I'd want the largest distance to be correct, and let any errors in the middle average themselves out, as there's not much you can do if your screws or rack are not consistent.

Having said that, it's unlikely that the calibration is the issue with your circle size.

If you cut a 6" pocket, is it off by the same amount?

First thing I'd do is cut a single, shallow slot and measure the width, to see if the tool is the correct size.

Then, cut two pockets, one climb cutting, and one conventional. See if there's a size difference. There usually will be.


Further I'm believing the CV Mode setting relates to circular or arc movements only and as noted above it is happening with both circular cuts and straight line movement.

No, CV mode affects all movement, not just arcs. But it should have no effect on size compared to Exact Stop mode.

pete1961
03-06-2014, 10:25 PM
Gecko G540, CNCRP Supplied along with the Nema23 Motors they sell.

pete1961
03-06-2014, 10:37 PM
I would definitely like to calibrate over a much longer distance but not sure what machinists tool I can use that will get me out to 24" that doesn't cost a fortune! The distances everything is off at the moment I'm not even close to at 5" to believe it would improve at say 24".

I did cut what was supposed to be a .750" X 24.00" Slot .563" Deep. Cut 3 of them and everyone was consistently off with the width being .810" X Depth being .610".
I will do some more test cutting to see if the error's are proportional. They do somewhat but not completely. Interesting that everything comes up short and not to long.

ger21
03-07-2014, 08:06 AM
If the depth is wrong, then I'd say that most likely the Z axis steps/unit are wrong.
Again, you need to cut some slots/pockets in the opposite direction to see if you're getting deflection.

As for measuring long distances, clamp a tape measure to your table and put a pointed bit in thwe collet. Try to get the point at the edge of a mark on the tape. You should be able to get pretty close this way.

Hack
03-07-2014, 08:06 PM
Silly question, but it your router bit new or has it been resharpened multiple times? Perhaps the bit is quite a bit undersize for the diameter you are using to program with? If you are using a 1/2" bit that has ben resharpened to .480, you now have a pocket that is .020 undersize. Doubt that is the problem, but worth a check.

Dan

pete1961
03-07-2014, 09:43 PM
Well for calibration purposes I use a drill rod in a precise bits collet. I did further checking this evening with a 12" electronic caliper and both the X and Y are out of callibration .006-.008". After multiple calibrations to just verify that my measuring is correct the results are the same. Changed the step pulse width and direction width and it had no effect. Diassembled the X Axis motor assembly from the rack to verify belt tension, pulley set screws, etc are tight and correct. They are all perfect. My initial belief of this being a mach3/G540/motor issue still holds. I find nothing mechanically wrong even rolled the X Axis on the V Con rails when I had the motor off and the movement is very smooth and noplay is felt in any of the mechanics of the gantry/carriage assembly.

neilw20
03-07-2014, 10:01 PM
Put in a pointed tool.
Move way from you reference to say1,1, using MDI, G1 X1 Y1 F200.
Line up some mark on a ruler with the point, using a magnifying glass to get it perfect without doing any jogging.
MDI X0 Y0
then MDI X1 Y1. Same place with magnifying glass? Should be.
now X2 Y2
and X1 Y1. Should return to EXACTLY the mark.
Do you have backlash? More than likely. A mechanical problem.
Back to X0 Y0
then X1 Y1. Surprised?
If a cutter has runout it will certainly cut oversize.

The only time to blame Mach3 is if you have backlash compensation on and set incorrectly or you are missing steps because motor tuning is not good.
And it still not the fault of Mach3. Just the setup or the hardware.
Does it have active high clock instead of active low or vice versa.
Important to have have correct clock phase because direction must not change at same time as clock(edge)
This is dependent on the driver setup and breakout configuration.


Set up a loop program:

X1 Y1
X0 Y0
X1 Y1
X0 Y0
X1 Y1
X0 Y0
X1 Y1
X0 Y0
X1 Y1
X0 Y0
X1 Y1
X0 Y0
X1 Y1

Should be the same place.
X2 Y2
X1 Y1
X2 Y2
X1 Y1
X2 Y2
X1 Y1
X2 Y2
X1 Y1
X2 Y2
X1 Y1
X2 Y2
X1 Y1
Same place? complete with error? Identify and fix the mechanical problem.

You can certainly have backlash, and not 'feel' it.
Cut some circles. Ar they PERFECTLY round or have flats (albeit small but usually visible) at the 90 degree points?

pete1961
03-07-2014, 10:34 PM
Could you explain the clock phase in a little more detail.

I will run the test you mention tommorow. If there is backlash it seems odd the backlash is exactly the same in either the X or Y direction.
I agree I definitely would not "feel" backlash by touching the machine.

neilw20
03-08-2014, 03:49 AM
Clock phase?
That information varies and is in the documentation for the controller.
There will be some waveforms there, and it relates to this. Often can be the cause of lost steps.

pete1961
03-08-2014, 09:38 AM
This morning did some move similar to what you mention but only in the X direction rather than diagonally and the results to me anyway are odd. Tell me if this is backlash or something else.

Moving away from 0,0
X0,Y0 =0.000
X1,Y0=.997
X2,Y0=2.000
X3,Y0=3.001
X4,Y0=4.006
X5,Y0=5.006
X6,Y0=6.000
X7,Y0=7.001
X8,Y0=7.998
X9,Y0=8.995
X10,Y0=10.000
X11,Y0=11.004
X12,Y0=12.003

Moving back to 0,0 as follows.
X11,Y0=11.013
X10,Y0=10.011
X9,Y0=9.005
X8,Y0=8.006
X7,Y0=7.010
X6,Y0=6.008
X5,Y0=5.013
X4,Y0=4.017
X3,Y0=3.011
X2,Y0=2.010
X1,Y0=1.008
X0,Y0=0.0

Ran these inputs 3 times out and back, same sequence, same machine start point, same exact measurements every time.

Ran another set of measurements from a different location on the table different results...

X0,Y0=0.000
X1,Y0=.989
X2,Y0=1.978
X3,Y0=2.974
X4,Y0=3.970
X5,Y0=4.965
X6,Y0=5.970
X7,Y0=6.977
X8,Y0=7.976
X9,Y0=8.982
X10,Y0=9.988
X11,Y0=10.989
X12,Y0=11.996

X11,Y0=10.999
X10,Y0=9.998
X9,Y0=8.994
X8,Y0=7.986
X7,Y0=6.986
X6,Y0=5.983
X5,Y0=4.976
X4,Y0=3.982
X3,Y0=2.989
X2,Y0=1.989
X1,Y0=1.000
X0,Y0=0.011

I have little doubt about my measurements being exact to .001"
If there is backlash wouldn't it be more consistent with each 1" move?
When I spoke to CNC Router Parts if I remember correctly there is virtually no backlash on the PRO rack unless someone can tell me otherwise.

Anyone have any further thoughts on this issue?

ger21
03-08-2014, 11:00 AM
Looks to me like you have .01" of backlash, ±.001-.002

pete1961
03-08-2014, 11:47 AM
Thanks for the reply, I have gone into mach3 to account for backlash.

Changing backlash settings has no effect on this problem tried .005", .008", & .010". Actually any of these numbers made no change at all.

I re-checked so mechanical issues and cut the same part again same errors. Measuring circles shows them to be off further in the "Y" direction than the "X" direction. Cutting a square yields worse results. Square should be 4.406" and machine cuts to 4.353" on the "Y" Axis. Didn't check the "X".

I'm going to also be scrutinizing the "Z" Axis as it appears to be cutting pockets deeper than my drawing shows them to be! They should be .193" machine is cutting them to .244"

I wonder if anyone else using a cncrp router can advise if this is common?

ahren
03-08-2014, 07:39 PM
Hi Tom,

Some of the issues could be related to the router not being trammed to the z axis. Have you surfaced your spoilboard yet? If not, this could also be impacting things. It's also possible that there is some looseness in the screws holding the angular contact bearings to the z-axis, and if so we can help you with this on Monday. All that being said, I'd hold off on dialing things in too much until you get the new parts you purchased, as you'll have to re-do a lot of it otherwise.

Best regards,

Ahren
CNCRouterParts (http://www.cncrouterparts.com)

neilw20
03-08-2014, 08:55 PM
Where you do
X0
X1
etc
make sure you get to initial X0 from a X- location as it appears you change direction at the start of the test.
Now you have figured the backlash is approx 0.01" set it going in a long loop..

X0
X0.01
X0
X0.01
...
X0
X0.01

and see if you can find the lost travel.
Maybe, with machine not moving, use a dial indicator and push/pull on axis by hand.

Best to fix the mechanical problem before trying to get rid of it by tuning Mach3.
Any backlash compensation causes all other axis but the compensated one to pause when there is a change in direction.
This means the tool will rub shiny spots in the job, like at the quadrant points of circles.

pete1961
03-08-2014, 11:15 PM
I'm with you Ahren will wait till the new parts are here and then work on the mechanics.
Thought I would get a better handle on the problem before the parts arrived.
At least I will be learning something.

FoxCNC1
03-09-2014, 06:40 PM
Hi Tom,

Some of the issues could be related to the router not being trammed to the z axis. Have you surfaced your spoilboard yet? If not, this could also be impacting things. It's also possible that there is some looseness in the screws holding the angular contact bearings to the z-axis, and if so we can help you with this on Monday. All that being said, I'd hold off on dialing things in too much until you get the new parts you purchased, as you'll have to re-do a lot of it otherwise.

Best regards,

Ahren
CNCRouterParts (http://www.cncrouterparts.com)


My machine was running baddddly... but I spend some long hours doing what Ahren mentioned and its now much better.
I think your problem is different than mine, but I did too have some issues with accuracy. I don't know how to check for back lash or how to compransate for it but, one thing that helped make things run better at least for me (NEMA 34) was to turn down the acceleration from 30.

pete1961
03-09-2014, 11:09 PM
I'm in the process of partially dismantling mine because I'm going to upgrade some parts to make the gantry taller and stiffer. While i was working on it tonight I did find slop in the carriage for the Z Axis that was noticble once I removed the Z axis unit from the carriage. This may account for some or all of the issue. I don't believe I have much of a backlash problem if any. The measurements are not consistent enough for me to think it has that much backlash and only when I move certain times. Cutting circles are so far off it is most likely something else. I will re-evaluate everything very closely when I re-assemble everything. I have quite of measuring devices that were used previously and will be used again to check everything. I will add to this post once I reach that point.

knightsofnii
04-24-2014, 04:25 PM
my biggest accuracy problem was solved when i realized one or both of my x axis motors seemed to be "hiccupping".

I pulled the drive components off, and I discovered that the tiny setscrews for the drive gears had spun out! two of them hit the floor and i lost one.
I went to my machine shop and got some more, and i put them in with some blue loctite and made sure they were as tight as possible.

After that, my accuracy increased tremendously!! I still have to do the Y axis, and anywhere else there are setscrews.
Just thought I'd mention to check these setscrews.

FoxCNC1
04-24-2014, 04:55 PM
Good point. I have had one of mine fall out.

pete1961
04-24-2014, 07:54 PM
I had used thread lock from the beginning. It was apparent to me that it was necessary when I noticed the motor shafts didn't have flats on them.

knightsofnii
04-24-2014, 10:58 PM
i also used setscrews that were a pinch longer and had some knurling on the tip.
If your shafts dont have a flat section, perhaps check that this part hasnt weakened?

FoxCNC1
04-24-2014, 11:15 PM
mine do.

fortyeight
05-14-2014, 07:23 PM
not concerning your topic.
But i was wondering about these kits. i am fairly new to cnc i have a small home built kit and lately all i have been at is looking at this kit pro4828.
was wondering exactly what its cutting size is, is it 48 by 24. how was the build process and the how unit working out.
Would appreciate all i can find out.

FoxCNC1
05-14-2014, 07:54 PM
The units works well. Just you have to be very methodical with the build etc - but that is the same for all machines.

fortyeight
05-14-2014, 08:04 PM
thanks. i would imagine so, taking time and care with assembly will probably make life easier when up and running.
And regarding it cutting size, is it a full 48 by 24